codes for the paper Assessing Dengue Forecasting Methods: A Comparative Study of Statistical Models and Machine Learning Techniques in Rio de Janeiro, Brazil

You can find the pre-print version of the paper here.

There are 2 parts of the models: first is using the cases itself (no-cov); the other is including covariates (cov).

The data is in data.csv, only including time and the dengue cases.

Main function is testing.R.

All the models are in the predict_functions.R.

Using ar_prediction_result <- predict_AR(data, window_size) can get a table of results including the real cases and predicting cases.

Then using print(combine_metrics(ar_prediction_result)) you can get a table of all 3 metrics of the model: MAE, MAPE, and RMSE.

The data is stored in data_with_covarites.csv, including time, cases, humidity and temperature.

The main function is testing.R, you can call sarimax_prediction_result <- predict_sarimax(data, window_size) to get the same result table of real cases and predicting cases, the same as the no-cov.

Then using the same function print(combine_metrics(sarimax_prediction_result)) you can get the metrics of MAE, MAPE, and RMSE.